 |
| UC San Diego estimator scientific discipline PhD candidate Oscar Beijbom |
Many students who consummate a masters or PhD at UC San Diego Jacobs School of Engineering volition locomote along to locomote entrepreneurs, launch startups, advance cutting-edge technology scientific discipline research, together with pioneer novel frontiers for the plain of engineering. Research Expo provides a unique chance to network amongst over 210 engineers of the futurity inwards i house together with time.
We’re taking this chance to innovate you lot to a few graduate students at the Jacobs School inwards advance of the final result on Apr 16, 2015. To start, nosotros sat downward amongst Oscar to larn to a greater extent than most his question together with the comport on he hopes it volition take away maintain inwards the future.
Q: Why UC San Diego?
A: I came to UC San Diego for my PhD to explore the interactions betwixt machine learning together with conservation together with sustainability. I’ve e'er been intrigued yesteryear the mightiness nosotros take away maintain to learn computers to exercise useful things. For example, earlier joining UC San Diego, I used machine learning to learn a estimator to recognize the motions of falling off of a wheel together with used that inwards plough to create an airbag bike-helmet. Check out that projection here.
Q: What did you lot receive upwards one's hear to report when you lot arrived, together with why?
A: I was given ii choices for my projection when I commencement arrived – coral reef ecology or aspect upwards recognition for the Navy. While both projects sounded exciting, it was a no-brainer for me to pick out coral reef ecology.
Q: Why’s that?
A: UC San Diego is unique inwards that it has Scripps Institution of Oceanography (SIO), a world-leader inwards ocean-science, conservation together with sustainability. I was glad to endure given the chance to endure a piece of work of the collaboration betwixt the Jacobs School of Engineering together with Scripps.
Q: What’s novel inwards coral reef ecology?
A: Climate alter stressors such equally ocean-acidification are causing the refuse of coral reef ecosystems all over the world, together with it is imperative that determination makers take away maintain accurate data. Thanks to advances inwards the speed of information collection such equally the proliferation of underwater autonomous or remotely operated vehicles, millions of images are collected every year. The employment is that researchers pass also much fourth dimension annotating coral reef images inwards club to position the organisms together with substrates that are introduce inwards each image.
Q: How tin technology scientific discipline solve that problem?
A: Under the supervision of my PhD advisor, professor of estimator scientific discipline together with technology scientific discipline David Kriegman, nosotros created an algorithm that annotates the images together with learns equally it goes. The algorithm relies on color together with texture descriptors. We take away maintain made it available to coral reef researchers on our site, CoralNet.
 |
| Coral reef survey image Source: CoralNet |
A: The site allows researchers to upload their survey images together with annotate them inwards the web-browser. On the dorsum end, a estimator learns from the annotations together with suggests annotations for the remaining images.
Q: Why is it together with thus of import that researchers endure able to trim the fourth dimension it takes to annotate these images?
A: Manual musical note is the bottleneck betwixt the fourth dimension it takes to capture images, which tin endure done quickly, together with the ecological information that decision-makers such equally reef managers or politicians exercise to implement critical changes inwards reef-management, marine-protected-areas, urban planning, agricultural runoff, etc.
Q: What comport on exercise you lot come across your question having on the means coral reef ecology is done inwards the future?
A: The algorithm is capable of reducing manual show on the average reef survey yesteryear fifty percent. This tin endure quite pregnant both inwards damage of cost-savings an inwards damage of enabling larger-scale studies. For example, the Coral Reef Ecosystem Division of the National Oceanic together with Atmospheric Administration (NOAA) spends roughly $100,000 a twelvemonth annotating these images. That translates to a $50,000 toll reduction using machine learning.
Oscar volition introduce a poster at Research Expo 2015 titled, “Automated musical note of coral reef survey images”.
Have nosotros peaked your interest? Visit http://www.jacobsschool.ucsd.edu/re/ to come across the posters together with programme together with to register.